基于分布式共享内存的RTI通信机制研究  

Research of RTI Communication Mechanism Based on Distributed Shared Memory

作　　者：梁洪波[1,2] 朱卫国[1] 姚益平[2] 刘步权[2] 慕晓冬[3] 李猛[2] 

机构地区：[1]中国人民解放军92117部队 [2]国防科技大学计算机学院 [3]第二炮兵工程大学

出　　处：《系统仿真学报》2013年第7期1484-1489,共6页Journal of System Simulation

基　　金：国家自然科学基金(61170047)

摘　　要：共享内存可以为运行于多核计算平台的进程间提供高效的通信方式,目前的HLA仿真运行支撑平台RTI主要采用基于TCP/IP的通信机制,不能充分发挥多核计算平台的核间高速共享内存通信优势。结合RTI体系结构和多核的通信特点,提出了一种基于分布式共享内存的RTI通信机制——SMCM机制,该机制为每个盟员设立一个独立的消息接收共享内存区域,发送盟员通过向接收盟员内存区写数据实现消息发送,并基于信号量机制实现盟员进程间访问控制,避免了共享内存访问冲突,提高了盟员间的通信效率。多核平台上的对象管理服务性能测试表明,与当前国际上普遍使用的pRTI相比,使用SMCM机制的RTI端对端通信延迟约为pRTI的1/3,吞吐量也明显高于pRTI。Within multi-core computing platform, high efficient communication between processes can be achieved by shared memory. However, most of current RTI software only provides TCP/IP communication support, and the advantages of shared memory communication cannot be fully taken. Combining the characteristics of RTI architecture and multi-core communication, an RTI communication mechanism based on distributed shared memory was proposed. In this mechanism, an independent shared memory used for message receiving was created for each federate, and sending federate couM realize message sent by writing message into the memory of receiving federate. The semaphore mechanism was taken to realize the accessing control between federates, and the violation of memory accessing could be avoided. By the benefits of these mechanisms, synchronization cost between sending federates could be reduced, and the communication efficiency could be improved. Experiments result shows that, under the multi-core environment, compared with pRTI which is widely employed in HLA simulation, the attribute updating latency by this mechanism is one third of that in pRTI, and the attribute updating throughput is greater than pRTI.

关 键 词：分布式交互仿真 HLA仿真 运行支撑环境(RTI) 多核平台 共享内存通信 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]